OpenAI's DevDay on 29 September 2026 shipped dots, GPT-6.1 Sol, the Ultrafast speed tier, the Pro 500 plan, Codex Security Cloud, ChatGPT Space and Pages, and more than a dozen smaller things. On X it was eight posts from the company account across two days, plus five from Sam Altman.
This page reads them the way the Doomers launch directory reads every launch: views, rank, and what the sequence says. Figures were read on the afternoon of 30 September.
The posts, in order
| When | Post | Views | Rank of 777 |
|---|---|---|---|
| 28 Sep | "Get ready." with a teaser video | 17.0M | 10 |
| 29 Sep, morning | "10am PT, on the dot." | 2.8M | 70 |
| 29 Sep, 10am PT | "Your dot is ready to meet you." | 6.7M | 26 |
| 29 Sep | "Introducing dots, powered by GPT-6 Astra." | 12.1M | 16 |
| 29 Sep | "GPT-6.1 Sol: near-Astra intelligence for a fifth of the price." | 2.8M | 70 |
| 29 Sep | "This is Ultrafast." | 1.1M | 183 |
| 29 Sep | Codex Security Cloud upgrade | 492K | top 40% |
| 29 Sep | The detail replies under each announcement | 115K to 484K each | around the median |
Altman's own account ran alongside: "Pretty excited for DevDay tomorrow. We have found a new thing." did 2.0M the night before.
"Dots are here!" did 1.7M. "Ultrafast is so fast I do not ever want to go back" did 923K.
His posts are shorter, later and smaller than the company's, and they read as a founder reacting to his own launch, which is the right division of labour when the company account is this large.
The teaser beat the reveal
The largest post of the launch was two words and a video, posted the day before, with no product in it. 17.0M views, the tenth largest post in a directory of 777 launches.
The reveal that followed did 12.1M. The teaser outperformed the thing it was teasing by 40%.
That is the ranking model doing what it does. A post that makes people ask "what is it?" collects replies and quotes at a high rate, and X pays for that with reach. A post that answers the question collects likes and bookmarks, which the model values less.
The algorithm guide covers the weights. The practical reading is that a launch is two posts, and the first one is the cheaper of the two to get right.
The countdown post in between, "10am PT, on the dot," did 2.8M. It kept the account in feeds through the morning without spending the reveal, and it planted the name six hours before anyone knew what it meant.
Twenty launches, one that travelled
GPT-6.1 Sol is a real model launch. Near-Astra performance at a fifth of the price, with a 95% cache-read discount, is the kind of announcement that would carry a day on its own. On this day it did 2.8M views.
When OpenAI launched GPT-6 Astra alone on 3 September, the post did 137M, the largest in the directory. When GPT-6 Sol and Luna shipped from the developers account on 22 September, that post did 978K.
Attention on X does not add up across posts from the same account on the same day. It concentrates on one and leaves the rest a fraction. Dots took the day.
Sol, which would have been the headline any other week, got a fifth of dots' reach, and Ultrafast got a tenth. The keynote format forces this. A company that ships 20 things in one hour has chosen which one the market will remember, whether or not it meant to.
The directory is built on this observation, which is why we launch one company a day and why a startup with three features ready should ship them on three days.
Sixteen launch partners, and where they showed up
Dots launched with 16 partners, including Notion, Vercel and Cognition's Devin. Each of them had a reason to post, and a partner's post reaches an audience the launching company does not own.
OpenAI does not need that reach. A startup launching an integration with a company of that size does, and the partner's post is the whole distribution plan.
The version of this a smaller company can run: the launch partners are agreed before the date, each has a post and an asset ready, and they go out inside the first hour so the ranking model sees the conversation forming. Creator amplification is the same mechanism with people who are not partners.
Three things to copy
- Tease the day before with nothing in it. Two words, a short video, a date. The question is the content. It works at every account size because it is built on replies, and replies are the one thing a small account can earn as fast as a large one.
- Plant the name before the reveal. "On the dot" put the word in front of 2.8M people before the product existed publicly. Whatever your product is called, people should have seen the word once before they see the launch post.
- Reveal one thing. Everything else goes in the replies or on another day. OpenAI could afford to bury Sol. Nobody else can.
The domain story that ran alongside, with dot.com pointing at a rival's product, is on its own page, along with the naming check that would have caught it.
Questions we get asked
What did OpenAI announce at DevDay 2026?
More than 20 things, led by dots, always-on agents built on GPT-6 Astra; GPT-6.1 Sol, a cheaper model near Astra's level; Ultrafast, a premium speed tier at up to 300 tokens a second; the Pro 500 plan; a Codex Security Cloud upgrade; and ChatGPT Space and Pages for working alongside agents.
How many views did the OpenAI dots launch get?
The reveal post reached 12.1M views by the afternoon of 30 September 2026, the 16th largest launch post in the Doomers directory of 777. The teaser the day before reached 17.0M.
Why did the teaser get more views than the launch?
A post that raises a question earns replies and quotes, which the ranking model rewards with reach. A post that answers it earns likes and bookmarks, which it rewards less. Teasers routinely outrun reveals for that reason.
How does DevDay compare with the GPT-6 Astra launch?
Astra's single launch post on 3 September 2026 reached 137M views, the largest in the directory. DevDay's largest post reached 17.0M and its reveal 12.1M, with more than 20 announcements sharing the day.
Should a startup launch several products on one day?
No. Attention on X concentrates on one post per account per day and leaves the rest a fraction. Ship the strongest one, put the rest in the replies, and launch the next on another day.